Automatic Invigilation Using Computer Vision

Educational institutions determine students’ strengths and weaknesses through exams. Students find numerous ways to cheat in physical exams like exchanging their sheets, using hidden notes, getting good grades, fulfilling their parents’ expectations, and whatnot. Due to the physical limitations of human supervisors, typical invigilation methods cannot conduct successful exams while maintaining their integrity. An automated method based on computer vision to detect anomalous activities during exams is proposed in this study. This study centers around invigilating students’ suspicious behaviour during physical exams through closed-circuit television (CCTV) cameras. The proposed method uses You Only Look Once (YOLOv3) with residual networks as the backbone architecture to inspect cheating in exams. The obtained results show the credibility and efficiency of the proposed method. The experimental results are promising and demonstrate the invigilation of the students in the examination. In this work, achieve 88.03% accuracy for the detection of cheating in the classroom environment